The Goal

KT Corporation wanted to find a smart, sector-specific solution that would help them provide viewers with a semantic search mechanism allowing them to locate and access more IPTV content and to match it with their preferences and needs.

The Challenge

Since the launch of the service in 2007, the existing search mechanism allowed viewers to discover less than 10% of all available content (more than 70,000 VOD contents). KT needed the service to be able to:

• analyze content to provide more relevant search results;
• provide metadata enrichment based on web mining (director, actor/actress, synopsis…);
• use multi-genre classification based on Association Rule Classifier;
• apply topic extraction based on ontology (that represented video categories/genres and emotional lexicons) and synonyms (from WordNet).

The Solution: Semantic Search to Expose More Content

KT implemented a semantic search system based on Ontotext’s technology with the following features:

• consonant-based query completion (a solution specific to Korean hieroglyphs);
• semantic-based query expansion;
• natural language query analysis based on VOD metadata: title, genre, actor/actress, director, etc.;
• semantic search based on an ontology including emotional words, topics, etc.;
• results filtering by user profile (user’s service grade);
• results ranking by user intention and statistics about VOD contents.

KT used a number of GraphDB licenses as an RDF database and reasoning service (supporting full OWL2 RL inference) to guarantee the required performance and resilience parameters.

The following is the KT conceptual diagram of the Interactive TV service:

Why Choose Ontotext?

With the help of Ontotext’s technology, KT is now able to better understand their customers and answer their information needs with a more efficient content discovery mechanism.

KT provides semantic VOD search service for IPTV and various smart mobile devices using OWLIM (now GraphDB) as semantic repository. We found that BigOWLIM is suitable for our service, as it has shown good performance for its price compared with other products in the actual simulation tests. Best of all, OWLIM could be easily tested, installed and used thanks to open library, detailed guidelines, and lots of samples etc. By and large, our experience is that OWLIM is a very powerful and reliable high-performance semantic repository.”

Joo Won Sung, Senior Researcher, Central R&D Laboratory, KT.

Do you think this case resembles your particular needs?